Research On User Groups Identification Based On The Propagation Of Specific Events In Social Network

The rapid development of social media platforms represented by Weibo and We Chat provide great convenience for the propagation of information,and false and spam information propagate quickly through the propagation mode.Behind the harmful information are user groups that push information and manipulate public opinion.The user groups disrupte the social network environment and the other users in social network have the negative impact,so the study of user groups identification is of great significance for maintaining network environment and creating good online public opinion.Traditional user groups recognition made progress to a certain extent.It mainly includes two aspects: one is the extraction of features,which mainly includes user attribute features,user behavior features,and content features;the other is recognition algorithm,which mainly use machine learning algorithms such as Decision Tree,Naive Bayes,and Support Vector Machine mine target user groups.Although the user groups can be effectively mined,with the rapid development of social network,user groups become more and more diversified,resulting in poor precision of existing user groups recognition methods.Considering that user groups play both the role of audience and the role of peddler during the event propagation.The features as the audience can be concealed,but the features as the peddler cannot be concealed,so the paper presents the research on user groups identification based on the propagation of the specific events.The main research contents are as follows:(1)Starting from the specific events in which user groups participate in the construction of online public opinion,combined with the unipolar negative public opinion characteristic of specific events,the network public opinion propagation model based on specific events is proposed.The Lyapunov stability analysis and simulation experiments verify the public opinion propagation proposed in the paper.The model conforms to the real law of the propagation of public opinion on specific events.(2)Starting from the prediction of network public opinion,in view of the lack of effective index system and sensitivity of the current public opinion prediction model,combined with the research on the propagation of network public opinion on specific events,the network public opinion prediction index system and the SNPOP model are proposed;firstly,the multiple index of group,emotion and influence are introduced.Construct the public opinion prediction index system;then integrate the specific event public opinion propagation model to construct the public opinion prediction model based on the specific event;finally,the simulation results verify the feasibility and effectiveness of the public opinion prediction index and model proposed in the paper.(3)Starting from the control of network public opinion,combined with the public opinion prediction model of specific events,the multi-features of user groups in social network and the Multiclass Support Vector Machine recognition algorithm based on EBPP are proposed;the public opinion index is the external manifestation of network public opinion,and the identification of user groups that contribute to the public opinion index is the key to public opinion control.Therefore,first analyze the similarities and differences between user groups in social network from the public opinion index feature,implicit network feature,social behavior feature,and semantics feature;then use EBPP to set the Multi-class Support Vector Machine parameters.Turning to the problem of particle swarm searching for elastic balance points in elastic space,the Multi-class Support Vector Machine recognition algorithm based on EBPP is constructed;finally,the Multi-class Support Vector Machine recognition algorithm based on EBPP is used to effectively identify multi-user groups in social network.The paper conducts data analysis on the microblog data set generated by grabbing,and the experiment verify that the method proposed in the paper is feasible and effective.At the same time,through multiple sets of comparative experiments,it is verified that the method proposed in the paper improve precision.